🏠 Chicago Airbnb Hybrid Recommender System

📜 Overview

This project develops a hybrid recommender system for Chicago Airbnb listings using data from Inside Airbnb. The system recommends listings based on three main preferences:

• 👥 Guest count
• 💲 Price point
• 📝 Text comments

Each recommendation includes the location and distance to the nearest CTA subway station for enhanced accessibility.

📊 Datasets

📌 1. Chicago Airbnb Listings

• Contains 7,952 listings with attributes like amenities, price, and location.

📝 2. Chicago Airbnb Reviews

• Includes 417,795 reviews, used for sentiment analysis.

🚇 3. Chicago CTA Stations

• Contains 145 train stations, used to compute distances from Airbnb listings.

🛠️ Methodology

1️⃣ Exploratory Analysis

🔹 Data visualization and attribute relationship analysis.

2️⃣ Review Scoring

🔹 Sentiment analysis using VADER and machine learning models (Logistic Regression, Linear SVC, Naive Bayes) to understand guest feedback.

3️⃣ Listing Clustering

🔹 Hierarchical clustering of Airbnb listings based on name and description embeddings to identify characteristic patterns.

4️⃣ Recommender System

🔹 Build a recommender system uses embeddings and cosine similarity for personalized recommendations and user preferences.

📈 Reviews Scoring - Sentiment Analysis

✅ VADER sentiment scores benchmarked against: Logistic Regression, Linear SVC, and Naive Bayes.
📌 Sentiment classes: Positive, Neutral, and Negative.
📌 Final sentiment score computed using a weighted average approach.
📌 Key Steps:

• Clean the review dataset.
• Address dataset skewness (majority of listings have high ratings).
• Utilize sentiment analysis techniques such as VADER to evaluate text sentiment.
• Generate word clouds to visualize sentiment trends.
• Libraries Used: nltk, pandas, matplotlib, wordcloud

🔍 Clustering

✅ Listings were grouped into three clusters based on semantic similarities.
1️⃣ Vibrant Social Spaces: Near entertainment and nightlife.

2️⃣ Cultural and Scenic Attractions: Near museums, parks, and historical sites.

3️⃣ Accessible with Public Transportation: Easy access to CTA train stations.
📌 Key Steps:

• Consolidate data from various sources.
• Process and clean text-based listing information.
• Apply Agglomerative Clustering to group similar listings.
• Visualize clustering results.
• Libraries Used: pandas, numpy, sklearn, spacy, matplotlib, geopy

🔥 Recommender System

✅ Implement a content-based recommendation model.
📌 Key Steps:
1️⃣ User Input & Preferences: Collects guest count, price, amenities, and text input to extract relevant features.
2️⃣ Filtering Listings: Filters listings based on user preferences.
3️⃣ Enhancing Rankings: Uses a combined sentiment and cosine similarity score between listings.
4️⃣ Personalized Recommendations: Outputs top-k listings from different clusters.
5️⃣ Geospatial Visualization: Displays results using an interactive Folium map.

• Libraries Used: pandas, numpy, sklearn, folium, spacy, geopandas

🎯 Results

✅ Best sentiment analysis model: Linear SVC (92.45% accuracy).
✅ Clustering provided interpretable themes for better recommendations.
✅ Embedding-based similarity scoring improved personalized recommendations.

🚀 Future Work

🔹 Implement dynamic weighting for ranking.
🔹 Integrate transformer-based models for better NLP understanding.
🔹 Expand the system to include real-time updates and user feedback mechanisms.

🌎 Links

🔹 🏠 Chicago Airbnb Listings Map: 🔗 📍 Chicago Airbnb Listings Map
🔹 🚇 Chicago Train Stations Map: 🔗 📍 View Chicago CTA Map

👨‍💻 Author independently developed all aspects of the project including data preprocessing, analysis, model development, and report creation.